HUMAN ERROR QUANTIFICATION: DEVELOPING NARA FROM HEART
|
|
- Gilbert Collins
- 5 years ago
- Views:
Transcription
1 HUMAN ERROR QUANTIFICATION: DEVELOPING NARA FROM HEART by Jim Edmunds Presented at the UK Inaugural PSA/HFA Forum Held at Leatherhead on 9 September
2 Outline History the need for Human Error Quantification HEART what we use now NARA an update and improvement of HEART 2
3 History Early UK PSAs First PSAs for operational UK NPPs carried out in early/mid 1980s No systematic representation of operator interactions A few exceptional actions represented nominal reliabilities claimed 3
4 History Introducing HFA into PSAs Mid/late 80s increasing systematic representation of operator interactions, leading to the need for: - Systematic Human Factors Assessment (HFA), and - A method of Human Error Quantification (HEQ) HEART used for HEQ based on HEART - A Proposed Method for Assessing and Reducing Human Error (Williams, 1986) 4
5 History Maturing HFA 1990s fully integrated HFA/PSA, including extensive Task Analysis Operator action dependencies considered Benefit of extended time taken into account A User Manual for HEART (Williams, 1992) CORE-DATA human error database project commences HEART validated against THERP and JHEDI 5
6 Principles of HEART Data-based Quick and flexible Not requiring extensive decomposition Intentionally conservative/pessimistic Takes account of important performance shaping factors 6
7 Using HEART Select Generic Task Types Characterise the operator action being considered based on the data gathered during Task Analysis Select a Generic Task Type (GTT) to represent the action being quantified, e.g. GTT F, change plant state using procedures, nominal Human Error Probability (HEP) = 3.0x10-3 7
8 Using HEART Appling Error Producing Conditions Consider differences between ideal and real conditions, e.g. operator unfamiliarity Select one or more Error Producing Conditions (EPCs) to represent these differences, e.g. EPC 1, unfamiliarity, having a maximum effect of 17 Judge the strength of the affect of each chosen EPC on successful task performance and apply an Assessed Proportion of Affect (APOA), e.g. apply an APOA of 0.5 if an operator is trained to carry out an action but would be expected to perform it very rarely 8
9 Final HEART HEP Calculation HEP = GTTx[(EPC 1-1)xAPOA 1 ]x [(EPC n -1)xAPOA n ] Task Description: Operator fails to start pump Generic Task Type selected: F (change plant state using procedures) HEP = Error Producing Conditions Assessed Effect 2 - Time Pressure (x 11) ( ( 11-1) x 0.1) + 1 = Operator inexperience (x 3) ( ( 3-1) x 0.3) + 1 = 1.6 Final HEP = x 2 x 1.6 =
10 Disadvantages of HEART Considerable assessor judgement needed Dependency within a sequence of operator actions not dealt with Does not credit long timescales Based on old data, much of which not relevant to the nuclear power context Does not deal with violations or errors of commission 10
11 Background to NARA Development Over 10 years experience of HEART in UK NPP PSAs Regulator pressure to update data and improve transparency of GTT and EPC derivation Need to make the application of HEART more consistent Additional functionality required to cover: - Dependency - Extended time 11
12 Project Team Ian Umbers Jim Edmunds Barry Kirwan Huw Gibson Richard Kennedy Garry Cooksley/ Jasbir Sidhu HFA Specialist and BE Project Officer PSA/HFA Expert and CRA Project Manager HRA Expert and Technical Leader Human Error Data (CORE-DATA) Specialist HEART Development Specialist PSA/HFA Experts 12
13 NARA AIMS Specific to UK NPP application Transparent use of data Open to being updated Benefits of long timescales credited Dependency guidance given Consider errors of commission 13
14 NARA GTT Derivation Review of UK NPP PSAs led to a refinement of GTT definitions to more accurately encompass the actions being considered in the PSAs CORE-DATA used to quantify GTTs using a transparent data scoring methodology 14
15 NARA EPC Derivation Review of PSAs to identify EPCs used Reduced set of EPCs developed for the IMC Consistency in HRA project Additional EPCs were also generated from reviews of other approaches to human error identification Maximum affects derived by review and analysis of available original HEART data sources, and some use of CORE-DATA 15
16 Benefits of Extended Timescales Some scenarios in gas-cooled reactors offer extended timescales, from 3-24 hours Further attempts to complete a task can be made in this time, if first attempts have failed Almost no data on such long timescale events NARA has built on the prior IMC DORRET work on human error in long timescale events 16
17 Pre-Conditions for Crediting Extended Timescales Need for action remains clear Options for completing the action are apparent Enough people available to respond Increasing benefits can only be claimed with increasingly good key determining factors NARA provides a detailed checklist for claiming ETFs 17
18 Considering Dependencies For failure sequences which include multiple operator actions, dependencies must be considered Three approaches compatible with NARA - Dependency Modelling - Dependency Specific Limits - Human Performance Limiting Values 18
19 Recording the Information Used Records are needed to bring together all the information used in the final assessment This information should provide the rationale for choosing the GTTs, EPCs, APOAs, ETFs and any Dependency Factors which help determine the calculated HEPs 19
20 Overview of the Human Error Quantification Process Step 1: Identify and Describe Human Interactions to be Quantified Step 3: Select a Representative Generic Task Type (GTT) Step 2: Carry out Task Analysis (HTA, TTA, DAD, TLD) Step 4: Select & Proportion Error Producing Conditions (EPC) Step 4.1: Select Applicable EPC Step 4.2: Check for EPC Double Counting Step 4.3: Assign Assessed Proportion of Affect (APOA) Do Extended Time Factors (ETF) Apply? YES Step 5: Consider Extended Time Factors NO Step 6: Calculate Human Error Probability (HEP) Do Dependencies Apply? YES Step 7: Consider Dependencies NO Step 8: Record Rationale & Document Final HEP 20
21 Strengths of NARA Derivation of GTTs and EPCs fully open to scrutiny - a first for a data-based HRA approach This also means that NARA can evolve as additional data become available Particularly adapted to the needs of UK NPP PSAs Remains flexible and relatively rapid Additional functionality (dependency and long timescales) make the HRA process more coherent More guidance for the assessor 21
22 Limitations of NARA Validation - NARA s validity linked to HEART s, but may ultimately require proof - validations not cheap Data limitations in certain areas - some GTTs under-pinned by fewer than 10 data points Errors of commission - not yet operational 22
23 Future Developments Response to 2 nd Peer Review Regulator approval Promotion of technique Application to different industries Gaining acceptance of the theoretical basis Validation Automation 23
24 DEVELOPING NARA FROM HEART Thank you for listening Any questions? 24